Americans: Test your broadband speed, help the FCC keep ISPs honest

James from the New America foundation sez, "The FCC launched a consumer broadband test on their blog yesterday. Internet speeds in the US are often 50% to 80% lower than advertised and its vital consumers have reliable information on the actual performance of their connections. One of the two tools the FCC is using is the Network Diagnostic Tool (NDT), an open source tool hosted on (M-Lab). The validity of NDT can be independently verified, and all data is publicly released. M-Lab hosts other test as wells, such as a test to see if bit torrent is being throttled, or how much bandwidth is available."

Consumer Broadband Test (Thanks, James!)


  1. I’m already suspecting shenanigans.

    I did the test to the Ookla servers last night from my Time Warner account and got about 28000 kbps down on five separate occasions, which is nearly double my advertised 15 Mbit down speed. Later I did the connection to MLAB and got 8200 kbps, much more reasonable, though still slower than advertised. On both tests my upstream came out to around 730 kbps.

    Unless they’re using highly compressible data at Ookla, I can’t figure out how I’m getting numbers so high other than someone at Time Warner unlocking my bandwidth cap just for this test.

  2. I just switched last night from Verizon Cable to ATT DSL, so it was the perfect time to try this. Our connection speed seems noticeably faster, but I don’t know what the numbers mean, can anyone help out?

    The results of this test are: Download speed: 5092 kbps Upload Speed 646 kbps with 20 ms latency. I was told by ATT that our service is “HS Elite” good for 6 mbps, so I know the top number is close. But is this asynchronous? That does not bode well for my web postings if upload is 10x slower than download….

    I am hoping that the Google fiber cities 1 Gbps initiative will come in here to Ann Arbor and sweep us off our feet!

    1. Verizon does not offer cable. They do offer Fios which is capable of delivering TV signals but it is fiber optic and therefore not what one would traditionally think of as cable (despite what Comcast would like you to think).

      Most residential DSL connections are asynchronous or ADSL. The reason being that the large majority of residential customers are going to be doing much more downstream traffic than upstream.

    2. You are getting “normal” speed. I am subscribed to same plan and im getting similar numbers. The thing is, ATT’s advertised speed is 7Mb/s which is ~7100kbps give or take(compared to 5000 we get)

    3. You are lucky because you are getting right. I am paying to ATT for 6 and I am getting 2 or 3. Keep praying.

  3. This is the start of something good, but if I can’t put in what advertised speeds I am paying for and from whom, how is this keeping anyone honest? It’s just mapping connection quality over geographic area.

  4. If you are paying for a 6mbps connection then your getting pretty close to that.
    when running speed tests, your download speed will never be the full advertised speed your ISP says due to a few factors:
    1 – the connection at the other end of the line. If the server your connecting to is slow, then your connection is going to apear slow.
    2 – Amount of traffic your ISP is handleing. If their servers start to max out due to X amount of bandwidth alotted, then you will see slow down.
    There are other factors as well, such as your computer and network setup.

    As for your 646kbps upload. This is normal, most ISPs in North America will only give you 512kbps upload on a cable or DSL connection. As downstream increases, say with a 20mbps connection, ISP’s may boost up the upstream speed to 1mbps. But in most cases they keep with a 75/25 ratio.

    The main thing to remember is that when an ISP advertises its speed, they usually mean to their servers, after data passes their servers, they have very little control on the speed. It all depends on the guy at the other end.

  5. But ALL providers use the magical phrase “up to”, as in “up to 6mbps”. The only guarantee is that you won’t get *more* than X.

    1. You can complain to the FCC. Now, whether the FCC actually has the power to do anything about it is exactly what the net neutrality debate is all about.

  6. Re Agnes/Verizon – Oops I meant Comcast cable- I was looking at another screen comparing cellular plans in my never-ending quest to save money I am thinking about switching from Verizon.

  7. My ISP, Cox, uses a technology they call “SpeedBoost” that messes with download speed tests. Essentially, they cache the files locally, so it always shows ridiculously high speeds. For example, I only pay for 7Mbps, but if I go to, it shows I have 22-30Mbps, because Cox is caching the files on their servers. In reality, my maximum is still around 6-7 Mbps.

    1. Comcast has something called Speedboost and what it does is opens the “pipes” up a bit faster during the first couple minutes of a file download. So if you’re paying for 12 MB down, these speedtests should show speeds between 12 and 24MB. I’m currently a 12 MB customer and my test showed 14.6MB down and about 2.7 up, with about 37ms latency.

      1. Comcast’s advertised speeds are the “Powerboost” speeds.

        From Comcast’s site: “Get download speeds up to 15 Mbps and uploads up to 3 Mbps with PowerBoost After the first 10MB of a file are downloaded, Powerboost is turned off and you’re dropped to the default speed.

        Finding out what the default speed is tricky, since I’ve tried and no Comcast CSR has ever been able to tell me. For the 15 Mbps package I have I believe it to be actually 6 Mbps, since that was the rate prior to their introduction of Powerboost, and you can find some left-over verbiage deep in their website.

        As for my FCC test results, despite my 15/6 Mbps package I got a paltry 2 Mbps download rate. Ok. Now what? Regardless of the questionable validity of this or any other speedtest, I think Comcast has their asses pretty well covered by only saying “up to” when advertising broadband speeds.

  8. Consumers don’t understand the difference between bandwidth and throughput, they’re generally not going to eliminate any of the obvious overhead on their home networks.

    Likely, it will result in many people assuming their ISP is somehow screwing them because they’re never going to see exactly what they’re paying for.

  9. The moment the test finished, I was re-directed to Sun’s Java page (No, I do not want to download the new version right now, thank you), preventing me from seeing the last result. Also, a requirement for an address is pretty sucky since they can probably gather any geolocation info from my IP address. That all said, this is a good idea.

  10. Why does it need my address? I trust the Gub-mint less then I trust my ISP. I’ve checked my speed connection numerous times on websites without having to give an address.

  11. Doran, aren’t you aware of the fact that anyone you make a TCP/IP connection with knows your IP address and can look it up in a geolocation database? They really don’t need to ask for your address, they could retrieve it from the connection you made when you ran the test.

  12. If people like Mitch mess with the quality of the data, they’ll just do what they should have done in the first place and log the IP address used to run the performance test rather than asking for it, and put Mitch down on a list of liars who maybe should have their tax returns checked for honesty. OK, seriously, there’s no reason for them to ask the question and rely on the answer given.

  13. Very interesting project here…. Wonder how the tinfoil hat crowd is going to react to surrendering precise street address for a simple speedtest? (Conversely, how skewed will the dataset be when n% of testers provide a phony address?) How many BOFH are going to scream as business users try to DL/install/activate Java? (At least has the good sense to direct my Mac to Apple’s updater ;-)

    Presentation seems a bit beta… “Users are randomly assigned the Ookla or M-Lab application. Note: the M-Lab application currently does not work with Safari, Chrome, and Opera web browsers.” So I’ve gotta drill through a buncha links to actually select the Ookla applet; they can’t just read my UA string and send me to the correct app? There’s a privacy policy on the main page, but no “accept/reject” selector; no disclaimer about running the applet. This kind of stuff would get a commercial site skewered in the punditverse, no?

    I’ve got no desire to jump through these hoops, so I’ll just stick with the “traditional” speed test over at; for latency I’ll just throw some pings at my ISP, my ISP’s competitor, and

  14. “Users are randomly assigned the Ookla or M-Lab application.
    Note: the M-Lab application currently does not work with Safari, Chrome, and Opera web browsers.”

    So are they randomly assigning Chrome users to M-Lab or are they not actually randomly assigning you?

  15. @kilranian – ugh we had to get Cox to stop doing that here locally, as it was affecting our University and proxy access for journals and databases.

  16. Running bandwidth test, from a .gov site, that tests my system by upping and downing a file to my machine right?

    I am sure the benign file they test with is legit, but no thanks. I don’t have a tinfoil hat but I am sticking to speakeasy as well.

  17. Problem: ISPs can set up special routing so that this broadband test gets higher priority and preferential treatment as opposed to normal traffic. Unless the FCC is auditing ISPs at random looking for this behavior, all of this is a total failure.

    The only worthwhile speed test is to use your internet for a long period of time, both burst transfers and longer, large downloads.

  18. What are the legalities regarding the movie picture association trying to convince the FCC to allow them to release first run movies directly to the consumer instead of releasing the material to theaters? Is that even legal?

  19. This government speed test is a joke. I tested by broadband connection with several different java and flash speed test tools. Most of my results came in around 5,000 kbps download speed, which is what I expected and is what my provider (Time Warner) says it provides. When I take the FCC test above, it indicates my download speed is 15,000 kbps, which is not accurate. This test is not accurate.

  20. I balked because of the address field, and because I already know how to determine my bandwidth speed, although I am interested in what the FCC uses to test that.

    Shouldn’t they be able to get a sample of locale from IP address? (Except for those using dial up. )

    Would zip code not give them enough info?

    So, by asking for address, they skew the likelihood of a decent sample.

  21. I just tried the test twice and got no results at all. I’m on Comcast.

    Mimi-rant follows:

    The word your is the possessive case of the word you. The elided word you’re is a contraction of the two words you and are. There is a vast gulf in the meaning of these two words/phrases.

    I know this may be nitpicking but, don’t we all want to say what we mean?

    1. Mimi is someone’s name, yet you’re profile says you’re name is Rick. There’s a vast gulf of difference between these two words.

  22. Is it just me or am I the only one that doesn’t feel comfortable giving the government my IP address AND my physical address?

  23. I prefer the NDTs (Network Diagnostic Tools) myself. They only run two ten-second tests (one for upload and one for download speed), but they give a wealth of information about your connection if you click the “More Details” button. Java is required for the following two:

    This one is available for download as Source (tarball). More information can be found at: It is a command-line utility:

    1. The Stanford site gives me this interesting (albeit uncopypastable) info:

      The theoretical network limit is 317.28 Mbps.
      The NDT server has a 127.0 Kbyte buffer which limits the throughput to 56.96 Mbps.
      Your PC workstation has a 63 Kbyte buffer which limits the throughput to 14.24 Mbps.
      The network based flow control limits the throughput to 19.35 Mbps.

      So it appears that my machine is the limiting factor.

      1. Yeah, I put a wireless card in my desktop about a week ago and this is the first time I’ve checked the diagnostics from the NDTs thoroughly, and apparently I now have a 16.0 KByte buffer which limits the throughput to 4.74 Mbps. The connection is a 6 Mbps connection and I share it with another person on a laptop, so i’m still getting about the best I can, but I’m going to have to see if there’s some way to change that or if it’s just the card itself. (Anon #33 here, btw).

  24. I used to do DSL tech support for Verizon and they pretty much only guarantee 80% of advertised speed. They will not even act on a complaint if it is within 20% of what you are supposed to get.

  25. Since I live at an address, have a driver’s license, phone number, passport, pay my taxes, etc., I’m fairly certain the gov knows where I live. Time Warner cable is more likely to sell my info than the gov, so I have no prob with the address question.

    Paraphrasing a commenter from last week – tin foil hats make you look fat.

  26. I suggest running both tests. Ookla said I had a 14mb/s dl and M-Lab said 8mb/s. Unless something magical happened in the minute between tests, I’d say that we can expect to see a couple of very different data sets.

  27. Download Speed: 15571 kbps
    Upload Speed: 3851 kbps
    Latency: 19 ms
    Jitter: 6 ms

    I’ve never had a problem with my internet connection speed but what is preventing Comcast from just increasing my speed when I access this site like they decrease speeds when users transfer files via bit torrent?

  28. Over on the NANOG (North American Network Operator’s Group) mailing list, the prevailing opinion is that the address is requested so that if you’re on a DSL line, they can compute the number of cable feet you are from the central office, which impacts the max possible throughput
    of a DSL. If you’re 500 feet from the CO, the top speed possible is a lot higher than if you’re at 17,950 cable feet and the spec says 18,000 max.

  29. Awesome. I am getting 1/3 of the speed that I am paying for, up and down, with 1/2 second latency times! I love you AT&T!

  30. what about broadband providers that block VOIP and HTTP traffic while they offer phone and webhosting services??

  31. Being bored this eve I ran both tests, more than once. Mlab’s results are consistently quite different (and not as good) from Ookla’s, with an avg difference of 6000 kbps on downloads (uploads are fairly similar), 90 ms latency and 90 ms jitter. Why such a variance?

  32. Anyone else creeped out by all the personal information this government site wants you to fork over?

    For the record, entering random characters for address and city, along with a random state and zip get you past the gate keeper.

  33. The first attempt at a test failed halfway through because of an overloaded server.

    The second attempt had the file cached so it assumed my speed was 999kb/s, which is far in excess of my bandwidth cap (I’m not allowed to have speeds that fast on this plan). Speed testing isn’t supposed to allow files like that to be cached to ruin subsequent tests.

    This is an amateurish test that’s just like the crummy ones that first cropped up on the web years ago and fell out of favor because their results were meaningless.

  34. As expected, other speeds peg my connection at its rated 1 megabit (it’s the middle of the night, so hey, I get full speed!) whereas this, um… cute little test thinks I have 999-1024kb/s upload and 800+ upload (I actually have about 1/3 a megabit upload at best).

    This is a godawful test, and I tried both of them. They both got it just as wrong.

  35. The only test that matters is between the desktop, or mobile device, and the edge router of the ISP. Does this test measure that connection? I ran a company in 2001 that did, ConnectCheck. Our Speedcheck tool measured the connection from the desktop to the edge router, the last router belonging to the ISP before their packets are handed over to the backbone carrier.

    The ISP has no control over the Internet backbone. Unless the FCC’s tool measures the last mile connection it is not measuring what the providers are selling.

  36. I can’t think of a good excuse for them to be collecting my address on an unencrypted site. Debate as to whether or not IP addresses are personally identifiable information aside, there’s no reason not to budget a lousy $100 to buy an SSL certificate for this project.

Comments are closed.